Content and Context: Two-Pronged Bootstrapped Learning for Regex-Formatted Entity Extraction

Authors: Stanley Simoes, Deepak P, Munu Sairamesh, Deepak Khemani, Sameep Mehta

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through an empirical evaluation over multiple real world document corpora, we illustrate the effectiveness of our approach. We perform our empirical evaluation on a variety of extraction tasks over multiple real-world document corpora as shown in Table 2.
Researcher Affiliation Collaboration Stanley Simoes Indian Institute of Technology Madras stanley@cse.iitm.ac.in Deepak P Queen s University Belfast deepaksp@acm.org Munu Sairamesh Indian Institute of Technology Madras musram@gmail.com Deepak Khemani Indian Institute of Technology Madras khemani@iitm.ac.in Sameep Mehta IBM Research India sameepmehta@in.ibm.com
Pseudocode Yes Algorithm 1 MATCH-SET-EXPANSION
Open Source Code Yes 4Source code available at https://github.com/stanleyts/ Content NContext
Open Datasets Yes The talk.politics.mideast and misc.forsale corpora are taken from the 20 Newsgroups dataset6, whereas the Enron corpus is a random subset of 100k documents from the Enron Email Dataset7. The Web KB corpus8 is another popular document dataset. 6http://qwone.com/ jason/20Newsgroups/ 7https://www.cs.cmu.edu/ ./enron/ 8http://www.cs.cmu.edu/afs/cs.cmu.edu/project/theo51/www/co-training/data/
Dataset Splits No The paper references various datasets but does not provide specific details on training, validation, and test dataset splits (e.g., percentages, sample counts, or explicit splitting methodology).
Hardware Specification No The paper does not provide specific hardware details (such as GPU/CPU models, processor types, or memory amounts) used for running its experiments. It only vaguely mentions 'server facilities' in the acknowledgments.
Software Dependencies No The paper describes algorithms and models (e.g., logistic regression, Levenshtein automaton) but does not provide specific software names with version numbers for replication.
Experiment Setup Yes Our method uses three parameters: d, num, and p. We set these to 4, 150, and 1%, unless otherwise stated. We separately study the performance of our method across variations in these parameters.